Support AI agent · Accessible channels

Accessible Support Channel AI Agent

Offer more than one way to reach a human: inventory the channels a customer can use, record the modes each one carries, and leave the conformity information to a named accessibility owner.

Typical delivery
4–6 weeks
Deployment
Your stack
Named owner
Mode coverage
After launch
Agent Care

What this agent does

Records the channel, never certifies it

In

  1. 01

    A channel opens, and the modes it carries are recorded beside it on the day it opens.

  2. 02

    A mode is missing, and the channel is marked single-mode rather than counted as complete.

Reason

  1. 03

    A help desk is a support service, and Annex I Section III(d) names it in the operative text.

  2. 04

    A duty reads where available, so the Act asks nothing of a channel a company does not run.

  3. 05

    A channel speaks through one sense only, and Section III(b) asks for more than one.

Decide

  1. 06

    A web or app route is tested as perceivable, operable, understandable and robust, or it is not.

  2. 07

    A certificate is offered for a service, and there is none to hold: Article 13 is self-declared.

Out

  1. 08

    A footer still points at the EU ODR platform, repealed on 20 July 2025, and it points nowhere.

  2. 09

    Execute write actions only inside the approval boundaries agreed during implementation.

Product statement

Inventory, mode coverage and the evidence behind both belong to the agent. Conformity, the Article 13 information and any notification to an authority belong to a named accessibility owner.

Example workflow

One channel, opening to sign-off

  1. 1Channel registeredPhone line, web chat, email alias, relay route, in-app form or a social inbox
  2. 2Modes recordedThe senses it reaches, the assistive technologies claimed and the day each was last tested
  3. 3Coverage assessedThe modes carried, the gaps read against Section III, the alternatives and confidence
  4. 4Controls appliedScope checks, sensory-channel checks, accessible-format checks and confidence threshold

No human action required

Stages 1 to 4 run unaided, and nothing is declared conformant at any of them — the agent is recording, and the accessibility owner lane opens at the confidence gate.

5DecisionBranches at the confidence threshold
High confidence

Goes to the named accessibility owner.

Low confidence

Adds an assistive-technology test first.

Accessibility owner review

The channel is held with the modes it carries, the gaps found and the evidence behind each.

Sign off · Add a mode · Send to specialist review

Signed off — by the named accessibility owner

  1. 6Channel register updatedOnly where write access and records policy allow it
  2. 7Outcome evaluatedMode coverage, evidence currency, owner corrections and what review found

Corrections

Each accessibility-owner correction is counted in the evaluation.

What should not run autonomously

Human approval stays in control

Outside the boundary — human approval required

8 items

  • Declaring a support channel conformant.
    • Signing the Article 13 conformity information.
    • Deciding the Act applies to a service.
    • Notifying a national authority of non-conformity.
  • Automation boundary

    Agent acts unaided

    • ✓Keep an inventory of the channels a customer can use to reach support.
    • ✓Record the modes each channel offers and when each was last tested.
    • ✓Check the accessibility notice is itself accessible.
    • ✓Flag a channel that has quietly become the only route.

    Nothing is declared conformant here; a named accessibility owner does that and signs it.

    • Judging whether an alternative mode is adequate.
    • Telling a regulator that a channel meets Section III.
    • Closing a channel a customer still depends on.
  • Changes to mode rules or scope settings.

Example output

One support channel, annotated

This serves a support team asked how a channel meets Annex I Section III, in an accessible format and with no certificate to point at; below is one channel as the agent leaves it.

Channel record · single channel

Illustrative example
Channel
Billing help desk
Modes carried
Voice, with a text relay route and a web form
Compatibility claim
Screen reader, retested
Evidence of record
Relay test log, 3 August 2026
Confidence
Held undeclared
Held for
The accessibility owner, by name

As received

Taken from the channel register and the test log — nothing on this side is written by the agent.

What the record holds

  • Relay test log
  • Channel register
  • Release notes

Why no declaration here

Conformity here is self-declared, and a named person declares it.

What the score decides

Below the configured threshold the channel gets an assistive-technology test first.

Action

Sign offAdd a modeSend to specialist review

Value

Where AI adds value

The same four claims, placed at the point in the workflow where each one applies.

Where the value lands

Value 01 – 04

Every channel on the list

From the register that names it

03Coverage

Record, then hand over

The media-entertainment accessibility-certification agent is broadcast captioning and audio-description conformity; this one is whether a customer can reach a human at all.

01Approved path

A help desk is in scope

Named in the operative annex text rather than a recital, and applying since 28 June 2025 — though a microenterprise providing services is exempt, on headcount and on turnover or balance sheet.

02Human review

What was checked, and not found

Checked across the Directive: no certification body, no registration, no filing and no third-party conformity assessment exists for a service. Article 13 is self-declaration — Annex V information in an accessible format, procedures for continued conformity, notification of the national authority — and CE marking attaches to products, not services. The ODR link duty fell with Regulation 524/2013 on 20 July 2025 with no successor; the ADR Directive obliges a trader to inform, not to use, excludes internal complaint handling, its revision scheduled from 20 September 2028.

04

Build an evidence trail

The channel, the modes it offers and the person who signed it off stay together.

Integrations

Typical integrations

Five system groups connect to the same agent. Which of them are in scope is decided in discovery.

  • Telephony and contact centre

    • Genesys
    • Five9
    • NICE CXone
    • Voice queues and IVR menus in use
  • Chat and messaging

    • Intercom
    • Drift
    • WhatsApp
    • In-app and web chat widgets
  • Relay and assistive routes

    • Text relay
    • video relay services
    • Sign-language interpretation routes

Agent

Accessible support channel coverage

  • Reads the register
  • Records the modes
  • Holds for the owner
  • Ticketing and self-service

    • Zendesk
    • Freshdesk
    • Gorgias
    • Help centres and public contact forms
  • Observability & evaluation

    • OpenTelemetry
    • Langfuse
    • Supported monitoring/evaluation sources

Integration availability depends on the client's existing systems and API access.

Agent controls

Six layers between a channel and a claim about it

Six doorways along one hall, the last the widest. What can pass is set out in the map below.

L6 · Outermost — last line of defence

Inward → L1 · closest to the model

L6Rollback / safe modeNarrow the agent to inventory only when evaluation or production signals degrade.Roll back
L5Version monitoringTrack model, prompt, mode rules and scope-configuration changes.Track
L4TraceabilityRecord the channel, the modes, the tests behind them and the owner read.Record
L3Owner sign-offHold each channel record for a named accessibility owner; the hold governs sign-off, not whether the channel is usable.Gate
L2Policy guardrailsTest each record against the configured mode and scope rules; a failure returns the record.Restrict
L1Confidence thresholdsRoute a thin mode claim to an assistive-technology test before the owner sees it.Require review
Model coreChannel recorded — the modes carried, the gaps found, the evidence and confidence
L1 – L2
Test whether a record may stand
L3
Leaves the sign-off to a named accessibility owner
L4 – L5
Keep the channel and the mode behind it
L6
Falls back to the accessible route when signals degrade

How Nestack evaluates it

Evaluate the whole record — not only the channel that came out clean.

Coverage runs the whole depth of the workflow, and every layer is cut by slice.

Surface — the channel a customer actually finds

  1. E1Final-output evaluationDid the record match the modes the channel actually carries?
  2. E2Step-level evaluationDid the agent read the right channel, the right test log and the live scope rules?
  3. E3Tool evaluationDid it read the correct channel and write the correct mode record?
  4. E4Confidence calibrationDo low-confidence records actually attract more owner corrections?
  5. E5Slice evaluationHow does coverage change across specific channel types?
  6. E6Business outcomeHow many channels needed a correction before the owner signed off?

Floor — the customer who cannot get through

Failure modes

Where each failure originates in the agent

Seven failure modes, each placed where a channel first stops working.

Agent lifecycle

Direction of processing →

  1. 01 · Retrieval

    1 mode

    • OY-03

      Channel missing from the register

      A route customers use is not on the list.

    Stage gathers

    The channels, the modes, the tests and the dates

  2. 02 · Reasoning

    2 modes

    • OY-04

      Relay route dead, still counted

      The route is listed, but nothing answers on it.

    • OY-06

      Service assumed out of scope

      A service inside Article 2 is read as outside.

    Stage proposes

    The modes carried, the gaps found and confidence

  3. 03 · Tool / write

    2 modes

    • OY-02

      New channel opened single-mode

      A channel launches carrying one mode only.

    • OY-05

      Evidence bound to wrong channel

      The test log is filed against another route.

    Stage writes

    Only where write access and approval policy allow it

  4. 04 · Output

    1 mode

    • OY-01

      Accessibility notice in one mode

      The notice arrives only in the mode it describes.

    Stage returns

    The record the owner signs and a customer feels

  5. 05 · Change / Version

    1 mode

    • OY-07

      Compatibility claim not retested

      A release ships and the claim is left standing.

    Stage tracks

    Model, prompt, mode rules and scope dates

  • Sev-1 · a channel declared without a test
  • Sev-2 · a dead mode still on the record
  • Sev-3 · signals degrade, record held back
See the complete Customer Support failure-mode catalogue

Affected slices

Voice-only desks absorb the misses

A channel-level mode-coverage figure can read clean while the voice-only help desks are carrying very nearly all of the gaps. Nestack reports the gap rate by channel, not only in total.

Slice performance — reported separately, not only in aggregate

Illustrative example
SliceFailure rateLiftLift vs. thresholdStatus
Voice-only help desks9.8%3.7×Review
Relay and interpretation routes7.0%2.6×Review
In-app chat widgets4.3%1.6×Watch
Published help centres2.3%0.9×Normal

Bar: gap-rate lift vs. help-centre baseline · scale 0–4.0× · tick marks the 2.0× review threshold

2 of 4 slices over threshold

Evidence-linked improvement

What one remaining route costs

The loop shuts when the channel that offered one mode only is a regression case. That suite is what the next channel opened is measured against.

Improvement cycle · five stages

Switchback — the path turns at Improve and returns at Learn

  1. 01

    Detect

    Mode gaps rise on one channel type.

  2. 02

    Diagnose

    The support line that works beautifully, for anyone who can use a telephone, is worked backwards until one cause is left standing.

  3. 03

    Improve

    Every change goes out numbered, with the channels that caused it attached.

  4. 04

    Verify

    Nothing ships while a single touched channel case is still red.

  5. 05

    Learn

    The case stays on, and the mode rules are rewritten alongside it.

  6. Learn → Detect

    The return edge. The next channel opened is read against a suite one case longer.

Typical build scope

Twelve workstreams across six weeks

The build scope read against the delivery timeline. Week structure follows the six-week plan — discovery, channels, mode recording, evaluation, integration, then production validation and handover.

Workstream

Week 1Week 2Week 3Week 4Week 5Week 6

  1. 01Channel-inventory and automation-boundary discovery.Week 1
  2. 02Telephony, chat and relay assessment.Week 1, Week 2
  3. 03Mode-coverage and assistive-technology claim mapping.Week 1, Week 2
  4. 04Channel ingestion and mode capture.Week 2
  5. 05Mode-coverage logic and evidence binding.Week 2, Week 3
  6. 06Confidence scoring and owner routing.Week 3
  7. 07Accessibility owner review workflow.Week 3, Week 4
  8. 08Channel-platform integration.Week 5
  9. 09Mode and channel cases.Week 4, Week 5
  10. 10Guardrails and fallback controls.Week 4
  11. 11Channel-trail instrumentation.Week 5, Week 6
  12. 12Deployment, documentation and Agent Care handover.Week 6

12 workstreams · 6 weeks · bar shows the weeks a workstream is active — several run in parallel

Final scope and sequence confirmed in discovery

Engagement tiers

What each tier includes

Rows are the capabilities named in each tier's scope. Higher tiers include everything below them.

Capability✓ in scope · — not at this tierPilotOne channel set, one marketProductionProduction channel platformsAdvancedMultiple regions / brands
Introduced at Pilot
Mode recording to your rules✓✓✓
Accessibility owner sign-off✓✓✓
Channel-inventory baseline✓✓✓
Introduced at Production
Reporting by channel—✓✓
Accessibility review workflow in your systems—✓✓
Approved write-back—✓✓
Channel-platform integration—✓✓
Introduced at Advanced
Multi-jurisdiction scope rules——✓
Multi-stage accessibility approvals——✓
Large channel estates——✓
Multi-channel mode controls——✓
Build priceFrom $5,000From $8,000Custom quote
Final build price
Confirmed after discovery based on integrations, workflow complexity, channel spread, approval controls and deployment requirements.
Separate from build
Build pricing is separate from recurring Agent Care, which covers managed monitoring, evaluations, incidents and verified improvements after launch.

What we need from you

What you bring, and what we build with it

Each input maps to a piece of build scope and a week in the delivery timeline.

  1. 01Your live channels and who answers on each

    Channel inventory and mode capture

    Week 1
  2. 02Representative contacts from each channel

    Channel ingestion, mode logic and the coverage baseline

    Week 2
  3. 03The services you provide and who owns accessibility

    Scope mapping, mode binding and the automation boundary

    Week 1
  4. 04Access to relevant APIs, feeds or exports

    Telephony, chat and relay assessment, then integration setup

    Week 2
  5. 05Channels you would not want audited

    Mode cases and the evaluation round

    Week 4
  6. 06What no channel may assume

    Confidence scoring, owner routing, guardrails and release controls

    Week 3
  7. 07A named accessibility owner who signs off

    Owner sign-off workflow, then pilot and production validation

    Weeks 5–6

Nothing else is required

Deployment, documentation and Agent Care handover are ours.

Delivery timeline

Four phases across six weeks

No column was widened here to look even, and the pair in week five is a pair that actually runs.

Phase

W1

W2

W3

W4

W5

W6

  1. DiscoveryW1
  2. BuildW2 – W3
  3. EvaluateW4 – W5
  4. Pilot & LaunchW5 – W6

Week focus

W1Channel discovery, scope binding and the automation boundary

W2Channel-platform integration and the coverage baseline

W3Mode-recording workflow, evidence logic and sign-off controls

W4Evaluation suite, mode cases and failure-mode testing

W5Platform integration, pilot channels and targeted corrections

W6One channel review run under the accessibility owner, then Agent Care handover

Reading the band
A bar runs across the weeks its own work is named in, and the week five overlap is real.
At the end of W6
Once the channel record validates, Agent Care assumes the agent.
Duration
Six-week plan shown · typical delivery 4–6 weeks depending on scope confirmed in discovery.

Next step · Customer Support AI agent

Build an accessible-channel agent around the channels your customers actually use.

Show us every way a customer can reach your support team. If your service is inside Article 2 and you run a help desk, then Annex I Section III(d) has applied to it since 28 June 2025. There is no certificate to buy, and the ODR link in your footer points nowhere.

Typical delivery
4–6 weeks
Deployment
Your support channels
Named owner
Mode coverage
After launch
Agent Care available

Nestack Agents · Accessible support channels

AGT-SUP-02 · Agent Care available after launch

Build an accessible-channel agent

Four questions so the call starts from your actual setup — then pick a time that suits you.

About you
Your channels

No obligation. We’ll map the workflow, name the automation boundary and tell you if an agent is the wrong answer.